mirror of https://github.com/yt-dlp/yt-dlp.git
[limelight] Add new extractor
This commit is contained in:
parent
85557f635a
commit
ef5acfe32d
|
@ -295,6 +295,11 @@ from .lifenews import (
|
|||
LifeNewsIE,
|
||||
LifeEmbedIE,
|
||||
)
|
||||
from .limelight import (
|
||||
LimeLightMediaIE,
|
||||
LimeLightChannelIE,
|
||||
LimeLightChannelListIE,
|
||||
)
|
||||
from .liveleak import LiveLeakIE
|
||||
from .livestream import (
|
||||
LivestreamIE,
|
||||
|
|
|
@ -0,0 +1,176 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
int_or_none,
|
||||
determine_ext,
|
||||
)
|
||||
|
||||
|
||||
class LimeLightBaseIE(InfoExtractor):
|
||||
|
||||
def get_playlist_service(self, id, method):
|
||||
return self._download_json(self.PLAYLIST_SERVICE_URL % (id, method), id)
|
||||
|
||||
def get_api(self, orgId, id, method):
|
||||
return self._download_json(self.API_URL % (orgId, id, method), id)
|
||||
|
||||
def process_data(self, mobileUrls, streams, properties):
|
||||
video_id = properties['media_id']
|
||||
formats = []
|
||||
|
||||
for mobileUrl in mobileUrls:
|
||||
if '.m3u8' in mobileUrl['mobileUrl']:
|
||||
formats.extend(self._extract_m3u8_formats(mobileUrl['mobileUrl'], video_id))
|
||||
else:
|
||||
formats.append({'url': mobileUrl['mobileUrl']})
|
||||
|
||||
for stream in streams:
|
||||
if '.f4m' in stream['url']:
|
||||
formats.extend(self._extract_f4m_formats(stream['url'], video_id))
|
||||
else:
|
||||
fmt = {
|
||||
'url': stream.get('url'),
|
||||
'abr': stream.get('audioBitRate'),
|
||||
'vbr': stream.get('videoBitRate'),
|
||||
'fps': stream.get('videoFrameRate'),
|
||||
'width': stream.get('videoWidthInPixels'),
|
||||
'height': stream.get('videoHeightInPixels'),
|
||||
'ext': determine_ext(stream.get('url'))
|
||||
}
|
||||
rtmp = re.search(r'^(?P<url>rtmp://[^/]+/(?P<app>.+))/(?P<playpath>mp4:.+)$', stream['url'])
|
||||
if rtmp:
|
||||
fmt.update({
|
||||
'url': rtmp.group('url'),
|
||||
'play_path': rtmp.group('playpath'),
|
||||
'app': rtmp.group('app'),
|
||||
})
|
||||
formats.append(fmt)
|
||||
|
||||
self._sort_formats(formats)
|
||||
|
||||
title = properties['title']
|
||||
description = properties.get('description')
|
||||
timestamp = properties.get('create_date')
|
||||
duration = int_or_none(properties.get('duration_in_milliseconds'))
|
||||
filesize = properties.get('total_storage_in_bytes')
|
||||
categories = [properties.get('category')]
|
||||
thumbnails = [{
|
||||
'url': thumbnail.get('url'),
|
||||
'width': int_or_none(thumbnail.get('width')),
|
||||
'height': int_or_none(thumbnail.get('height')),
|
||||
} for thumbnail in properties.get('thumbnails')]
|
||||
subtitles = {caption.get('language_code'): [{'url': caption.get('url')}] for caption in properties.get('captions')}
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': title,
|
||||
'description': description,
|
||||
'formats': formats,
|
||||
'timestamp': timestamp,
|
||||
'duration': duration,
|
||||
'filesize': filesize,
|
||||
'categories': categories,
|
||||
'thumbnails': thumbnails,
|
||||
'subtitles': subtitles,
|
||||
}
|
||||
|
||||
|
||||
class LimeLightMediaIE(LimeLightBaseIE):
|
||||
IE_NAME = 'limelight'
|
||||
_VALID_URL = r'http://link\.videoplatform\.limelight\.com/media/?.*mediaId=(?P<id>[a-z0-9]{32})'
|
||||
_TEST = {
|
||||
'url': 'http://link.videoplatform.limelight.com/media/?mediaId=3ffd040b522b4485b6d84effc750cd86',
|
||||
'md5': '3213605088be599705677ef785db6972',
|
||||
'info_dict': {
|
||||
'id': '3ffd040b522b4485b6d84effc750cd86',
|
||||
'ext': 'mp4',
|
||||
'title': 'HaP and the HB Prince Trailer',
|
||||
'description': 'As Harry Potter begins his 6th year at Hogwarts School of Witchcraft and Wizardry, he discovers an old book marked mysteriously "This book is the property of the Half-Blood Prince" and begins to learn more about Lord Voldemort\'s dark past.',
|
||||
'thumbnail': 're:^https?://.*\.jpeg$',
|
||||
'duration': 144230,
|
||||
'timestamp': 1244136834,
|
||||
"upload_date": "20090604",
|
||||
}
|
||||
}
|
||||
PLAYLIST_SERVICE_URL = 'http://production-ps.lvp.llnw.net/r/PlaylistService/media/%s/%s'
|
||||
API_URL = 'http://api.video.limelight.com/rest/organizations/%s/media/%s/%s.json'
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
|
||||
mobile_json_data = self.get_playlist_service(video_id, 'getMobilePlaylistByMediaId')
|
||||
pc_json_data = self.get_playlist_service(video_id, 'getPlaylistByMediaId')
|
||||
properties = self.get_api(pc_json_data['orgId'], video_id, 'properties')
|
||||
|
||||
return self.process_data(mobile_json_data['mediaList'][0]['mobileUrls'], pc_json_data['playlistItems'][0]['streams'], properties)
|
||||
|
||||
|
||||
class LimeLightChannelIE(LimeLightBaseIE):
|
||||
IE_NAME = 'limelight:channel'
|
||||
_VALID_URL = r'http://link\.videoplatform\.limelight\.com/media/?.*channelId=(?P<id>[a-z0-9]{32})'
|
||||
_TEST = {
|
||||
'url': 'http://link.videoplatform.limelight.com/media/?channelId=ab6a524c379342f9b23642917020c082',
|
||||
'info_dict': {
|
||||
'id': 'ab6a524c379342f9b23642917020c082',
|
||||
'title': 'Javascript Sample Code',
|
||||
},
|
||||
'playlist_mincount': 3,
|
||||
}
|
||||
PLAYLIST_SERVICE_URL = 'http://production-ps.lvp.llnw.net/r/PlaylistService/channel/%s/%s'
|
||||
API_URL = 'http://api.video.limelight.com/rest/organizations/%s/channels/%s/%s.json'
|
||||
|
||||
def _real_extract(self, url):
|
||||
channel_id = self._match_id(url)
|
||||
|
||||
mobile_json_data = self.get_playlist_service(channel_id, 'getMobilePlaylistWithNItemsByChannelId?begin=0&count=-1')
|
||||
pc_json_data = self.get_playlist_service(channel_id, 'getPlaylistByChannelId')
|
||||
medias = self.get_api(pc_json_data['orgId'], channel_id, 'media')
|
||||
|
||||
entries = []
|
||||
for i in range(len(medias['media_list'])):
|
||||
entries.append(self.process_data(mobile_json_data['mediaList'][i]['mobileUrls'], pc_json_data['playlistItems'][i]['streams'], medias['media_list'][i]))
|
||||
|
||||
return {
|
||||
'id': channel_id,
|
||||
'title': pc_json_data['title'],
|
||||
'entries': entries,
|
||||
'_type': 'playlist',
|
||||
}
|
||||
|
||||
|
||||
class LimeLightChannelListIE(LimeLightBaseIE):
|
||||
IE_NAME = 'limelight:channel_list'
|
||||
_VALID_URL = r'http://link\.videoplatform\.limelight\.com/media/?.*channelListId=(?P<id>[a-z0-9]{32})'
|
||||
_TEST = {
|
||||
'url': 'http://link.videoplatform.limelight.com/media/?channelListId=301b117890c4465c8179ede21fd92e2b',
|
||||
'info_dict': {
|
||||
'id': '301b117890c4465c8179ede21fd92e2b',
|
||||
'title': 'Website - Hero Player',
|
||||
},
|
||||
'playlist_mincount': 2,
|
||||
}
|
||||
PLAYLIST_SERVICE_URL = 'http://production-ps.lvp.llnw.net/r/PlaylistService/channel_list/%s/%s'
|
||||
|
||||
def _real_extract(self, url):
|
||||
channel_list_id = self._match_id(url)
|
||||
|
||||
json_data = self.get_playlist_service(channel_list_id, 'getMobileChannelListById')
|
||||
|
||||
entries = []
|
||||
for channel in json_data['channelList']:
|
||||
entries.append({
|
||||
'url': 'http://link.videoplatform.limelight.com/media/?channelId=%s' % channel['id'],
|
||||
'_type': 'url',
|
||||
'ie_key': 'LimeLightChannel',
|
||||
})
|
||||
|
||||
return {
|
||||
'id': channel_list_id,
|
||||
'title': json_data['title'],
|
||||
'entries': entries,
|
||||
'_type': 'playlist',
|
||||
}
|
Loading…
Reference in New Issue